fix(web,mobile): render Codex citations and artifact templates

[rynfar]

Codex emits file citations and artifact templates as markdown directives. Pylon
rendered the raw directive syntax, so a Codex answer that cited a file showed the
markup instead of a link, and artifact templates were unusable.

Adds three client-runtime modules — a directive parser, file-citation resolution,
and artifact-template presentation — and wires them into web ChatMarkdown and the
mobile thread feed. Citations become links into the file viewer; artifact templates
render as an affordance that appends the template's prompt to the composer.

Adopted from upstream pingdotgg/t3code#8584 (c1e70b5f8). Brings five new
dependencies (unified, remark-parse, mdast-util-directive, micromark-*).

Adaptation — five conflicts, three of them dependency artifacts. Upstream merged
this after #8793 and #8807, so its diff carries pieces of both:

• ThreadFeed keeps Pylon's renderer rather than upstream's Animated.View
  disclosure wrapper. THREAD_FEED_DISCLOSURE_* and disclosureEnteringEntryIds
  come from #8793 and do not exist in Pylon; only the artifact-template callback
  is added.
• MessagesTimeline keeps Pylon's local commandProgramName (line ~2466) instead of
  importing client-runtime/work-log/command-label, a subpath #8793 introduces.
• The onImageExpand prop upstream threads through ChatMarkdown belongs to #8807,
  which lands separately in fix(web): open agent images in the expanded preview #212. Dropped here so the two do not collide.
• Import blocks and Pylon's NOOP_OPEN_ATTACHMENT → NOOP_DOWNLOAD_ATTACHMENT
  rename are merged Pylon-first.

Verified: vp test run --dir packages/client-runtime for the three new modules
(45 passed); --dir apps/web ChatMarkdown, ChatView.logic,
MessagesTimeline.logic, markdown-clipboard (156 passed); typecheck clean on
@t3tools/client-runtime, @t3tools/web, and @t3tools/mobile.

Not exercised against a live Codex turn, on either client. Worth a real-client
pass before merge.

Reviewed and integrated with Claude Opus 5 in Claude Code.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.5 KiB	13.5 KiB	−19 B (−0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+6 B (+0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.6 KiB	6.6 KiB	−25 B (−0.4%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	57.2 KiB	57.1 KiB	−44 B (−0.1%)	66.4 KiB	✅
Codex	Live turn messages	10	9	−1 (−10.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+58 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+3 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: 5e3fd47 · PR result: 43c4aee · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.5 KiB
• Claude decoded thread snapshot: 110.2 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Adversarial review pass found two issues, both fixed in the follow-up commit:

1. Six dead imports (ChatAttachment, ChatFileAttachment, ChatImageAttachment, resolveAssetUrl, formatAttachmentSize, squashAtomCommandFailure) carried in from upstream's attachment work, which this branch does not adopt. Three were value imports that would have entered the Metro bundle. Removed.

2. Mobile "Copy message" copied raw directive markup (::artifact-template{…}, :codex-file-citation{…}) while the same message rendered as a card or link. Web already copies the rendered form via renderCodexDirectivesForCopy in MessagesTimeline.logic; mobile now does too. Upstream has the same gap, so this is a Pylon-side improvement rather than a port regression.

Re-verified: @t3tools/mobile and @t3tools/web typecheck clean, vp lint clean on the touched file.